GLM-5.2 Is The New Best Open Model
Zvi's AI Roundups · Zvi Mowshowitz · 2026-06-22
Zvi Mowshowitz conducts a full capabilities analysis of GLM-5.2, Z.ai's open-weights model, concluding it is the strongest open model available while remaining substantially behind closed frontier models and likely heavily distilled from Claude Opus.
Appears in
Extraction
Topics: open-weights-modelsai-benchmarkschinese-ai-labsmodel-distillationllm-capabilities
Claims
- GLM-5.2 is the strongest currently available open-weights language model, scoring around Opus 4.7 on traditional benchmarks.
- GLM-5.2 is substantially behind the absolute frontier of closed models like Opus 4.8 and GPT-5.5, despite being the best open model.
- GLM-5.2 is very likely heavily distilled from Claude Opus, evidenced by its strong prior that it is Claude, its distinct Claude voice, and its use of a Claude harness.
- GLM-5.2 occupies an awkward commercial niche—not cheap enough for bulk tasks nor strong enough for the hardest tasks compared to closed model alternatives at similar or lower cost.
- The GLM-5.2 release meaningfully updates the estimate of open-model progress but does not close the frontier gap, which remains substantial.
Key quotes
GLM-5.2 is still substantially behind the absolute frontier, although plausibly on the cost-benefit Pareto frontier. It seems closer to the frontier than previous efforts, including probably closer than DeepSeek R1 was during the DeepSeek moment.
The correct take is clearly some form of 'this model is dope, great job everyone, but not as dope as the hype might suggest.'
It would surprise me greatly if GLM-5.2 was not heavily distilled from Claude Opus. That does not invalidate the model, but it does mean two things. Distilled models tend to generalize poorly. They overperform on benchmarks and benchmark-like tasks, and on the most common tasks, and underperform on less common tasks.